There's been so much traffic here today that it's actually been a bit difficult to get in to write another post. And unfortunately it's all due to the Merck announcement. Some sites that have a long and distinguished history of drug discovery are set to be closed up as if they were so many redundant discount store locations.
And that shows you that no one in the business is thinking about these things - how hey, this site has really done a lot for us (or more than they should have, given their size), and maybe we should hold on to them. As past closings at other companies have shown, that's probably one of the last factors on the list, and most of the time it probably doesn't even come in at all.
What matters, I'd say, is what you'd think matters: the sheer accounting cost. How much does it cost to keep Facility X open? How much would it cost to close it? And how much would we save, compared to what we're giving up? It's that last part where the real arguing starts, because there are many people (I'm one, sometimes) that say that research cultures vary from place to place, and that some sites just seem to have a better history of discovery than others. They're not interchangeable.
But it's very hard to make that argument. This sort of thing doesn't show up in the financial statements, and it's hard to quantify. You also can't count on it, either, because some places will have a good run of many years, and then (for some reason) go flat. Despite what consultants will tell you, I don't think that anyone has figured out what exactly makes a research culture work. That makes fixing a broken one a tall order, and it also makes it very hard to raise one in the way that you want it. It's a combination of the individual people, their managers, the projects they get to work on, the experience that they have (or get) with success. . .all sorts of hard-to-deal-with variables.
It's not just in research institutes, either: why did Hungary produce its ferocious run of world-class scientists during the mid-20th century? Who wouldn't want to reproduce such a thing, or remake the old Bell Labs, or whatever your favorite success story might be. The fact that no one seems to be able to do this on demand argues strongly that no one knows how. And if no one knows how, no one's going to decide based on it, either. Sad.